Performance of Artificial Neural Network in Nowcasting Summer Monsoon Rainfall: A case Study

Hudnurkar, Shilpa; Rayavarapu, Neela

doi:10.1109/punecon.2018.8745413

Cited by 4 publications

(1 citation statement)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…All records were normalized using min-max normalization. A simple architecture of ANN with 2n+1 hidden neurons in a single hidden layer was employed for this purpose, where n is a number of inputs [40]. The dataset was divided into training and testing sets.…”

mentioning

confidence: 99%

On the performance analysis of rainfall prediction using mutual information with artificial neural network

Hudnurkar

Rayavarapu

2023

IJECE

View full text Add to dashboard Cite

Monsoon rainfall prediction over a small geographic region is indeed a challenging task. This paper uses monthly means of climate variables, namely air temperature (AT), sea surface temperature (SST), and sea level pressure (SLP) over the globe, to predict monthly and seasonal summer monsoon rainfall over the state of Maharashtra, India. Mutual information correlates the temperature and pressure from a grid of 10° longitude X 10° latitude with Maharashtra’s monthly rainfall time series. Based on the correlations, selected features over the respective latitude and longitudes are given as inputs to an artificial neural network. It was observed that AT and SLP could predict monthly monsoon rainfall with excellent accuracy. The performance of the test dataset was evaluated through mean absolute error; root mean square error, correlation coefficient, Nash Sutcliffe model efficiency coefficient, and maximum rainfall prediction capability of the network. The individual climate variable model for AT performed better in all evaluation parameters except maximum rainfall capability, where the combined model 2 with AT, SLP and SST as predictors outperformed. The SLP-only model’s performance was comparable to the AT-only model. The combined model 1 with AT and SLP as predictors was found better than the combined model 2.

show abstract

mentioning

confidence: 99%